U.S. patent application number 10/210489 was filed with the patent office on 2004-02-05 for storage system and method for providing consistent data modification information.
Invention is credited to Chron, Edward Gustav, Menon, Jaishankar Moothedath.
Application Number | 20040024973 10/210489 |
Document ID | / |
Family ID | 31187346 |
Filed Date | 2004-02-05 |
United States Patent
Application |
20040024973 |
Kind Code |
A1 |
Chron, Edward Gustav ; et
al. |
February 5, 2004 |
Storage system and method for providing consistent data
modification information
Abstract
A data storage system and method for providing consistent data
to multiple clients based on data modification information as
existing data is updated and new data is written to the system. The
information indicates the modification status of each data block
and identifies which data blocks have been modified during a
certain time interval. The clients may query and update the
modification information by submitting requests through a request
processor. The data modification information includes an Altered
Block Map that indicates block modification status and a Toggle
Block Map that identifies which blocks have been modified. The
system further includes a Modification Counter a Pending Reset
Counter for improved recognition and handling of the modified
data.
Inventors: |
Chron, Edward Gustav; (Santa
Clara, CA) ; Menon, Jaishankar Moothedath; (San Jose,
CA) |
Correspondence
Address: |
KHANH Q. TRAN
IBM CORPORATION, INTELLECTUAL PROPERTY LAW
DEPT. C4TA/J2B
650 HARRY ROAD
SAN JOSE
CA
95120-6099
US
|
Family ID: |
31187346 |
Appl. No.: |
10/210489 |
Filed: |
July 31, 2002 |
Current U.S.
Class: |
711/144 ;
711/141 |
Current CPC
Class: |
G06F 3/064 20130101;
G06F 3/0689 20130101; Y10S 707/99953 20130101; G06F 3/0619
20130101 |
Class at
Publication: |
711/144 ;
711/141 |
International
Class: |
G06F 012/00 |
Claims
What is claimed is:
1. A storage system for providing data consistency to multiple
clients, comprising: a facility for storing data and data
modification information; and a request processor for handling
requests for data from the clients, said request processor capable
of identifying data that has been modified during a given time
interval based on the modification information and updating the
modification information to ensure that the data accessed by the
clients will be consistent.
2. The system as in claim 1 wherein: the data is stored on the
facility as data blocks; and the data modification information
includes an Altered Block Map and a Toggle Block Map, the Altered
Block Map indicating the modification status of the data blocks and
the Toggle Block Map indicating which data blocks have been
modified.
3. The system as in claim 2 wherein: the Toggle Block Map includes
a plurality of bits, each bit corresponding to a data block; and
said bit is complemented when the data block is modified.
4. The system as in claim 2 wherein: the Altered Block Map includes
a plurality of bits, each pair of bits in the Altered Block Map
corresponding to a data block and including a Modification Bit and
a Pending Reset Bit; the Modification Bit is set whenever the data
block is modified by a write request; and the Pending Reset Bit is
set whenever the data block is requested by a client.
5. The system as in claim 4 wherein the Modification Bit and the
Pending Reset Bit may be reset by a request from one of the clients
to the request processor.
6. The system as in claim 5 wherein, if the data block is further
modified after the respective Pending Reset Bit has been set, then
the Pending Reset Bit is reset while the Modification Bit remains
set.
7. The system as in claim 5 further comprising an Altered Block
Counter for indicating the number of data blocks that have been
modified, the Altered Block Counter being incremented whenever one
of the data blocks is modified.
8. The system as in claim 7 wherein if the Modification Bit is
reset while the Pending Reset Bit is still on, then: the Altered
Block Counter is decremented; and the Pending Reset Bit is
reset.
9. The system as in claim 2 further comprising a set of application
programs wherein: the clients access the data blocks through the
application programs; and the application programs communicate with
the request processor to access the desired data blocks and to
update the modification information.
10. The system as in claim 2 further comprising a Modification
Counter, said Modification Counter being incremented every time one
of the data blocks is committed as a result of a client
request.
11. In a storage system including a facility for storing data and
data modification information and a request processor for handling
requests for data from clients, a method for providing data
consistency comprising the steps of: identifying data that has been
modified during a given time interval based on the modification
information; accessing the modified data; and updating the
modification information to ensure that data accessed by the
clients will be consistent.
12. The method as in claim 11 wherein: the data is stored on the
facility as data blocks; and the data modification information
includes a Toggle Block Map and an Altered Block Map, the Toggle
Block Map indicating which data blocks have been modified, and the
Altered Block Map indicating the modification status of the data
blocks.
13. The method as in claim 12 wherein: the Toggle Block Map
includes a plurality of bits, each bit corresponding to a data
block; and said bit is complemented when the data block is
modified.
14. The method as in claim 12 wherein: the Altered Block Map
includes a plurality of bits, each pair of bits in the Altered
Block Map corresponding to a data block and including a
Modification Bit and a Pending Reset Bit, the method further
comprising the steps of: setting the Modification Bit whenever the
data block is modified by a WRITE request; and setting the Pending
Reset Bit whenever the data block is requested by a client.
15. The method as in claim 14 further comprising the step of
resetting the Modification Bit and the Pending Reset Bit upon a
request from one of the clients to the request processor.
16. The method as in claim 15 further comprising the step of
resetting the Pending Reset Bit associated with the data block if
the data block is further modified after the respective Pending
Reset Bit has been set and while the Modification Bit remains
set.
17. The method as in claim 15 wherein: the storage system further
includes an Altered Block Counter for indicating the number of data
blocks that have been modified; and the method further comprises
the step of incrementing the Altered Block Counter whenever one of
the data blocks is modified.
18. The method as in claim 17 further comprising the step of
decrementing the Altered Block Counter whenever the Modification
Bit is reset.
19. The method as in claim 18 further comprising the steps of
decrementing the Altered Block Counter and resetting the Pending
Reset Bit whenever the Modification Bit is reset while the Pending
Reset Bit is still on.
20. The method as in claim 12 wherein: the clients access the data
blocks through a set of application programs; and the application
programs communicate with the request processor to access the
desired data blocks and to update the modification information.
21. A computer-program product for use with a storage system for
providing data consistency, the storage system including a facility
for storing data and data modification information and a request
processor for handling requests for data from clients, the
computer-program product comprising: a computer-readable medium;
means, provided on the computer-readable medium, for identifying
data that has been modified during a given time interval based on
the modification information; means, provided on the
computer-readable medium, for accessing the modified data; and
means, provided on the computer-readable medium, for updating the
modification information to ensure that data accessed by the
clients will be consistent.
22. The computer-program product as in claim 21, wherein: the data
is stored on the facility as data blocks; and the data modification
information includes a Toggle Block Map and an Altered Block Map,
the Toggle Block Map indicating which data blocks have been
modified, and the Altered Block Map indicating the modification
status of the data blocks.
23. The computer-program product as in claim 22 wherein: the Toggle
Block Map includes a plurality of bits, each bit corresponding to a
data block; and said bit is complemented when the data block is
modified.
24. The computer-program product as in claim 22 wherein: the
Altered Block Map includes a plurality of bits, each pair of bits
in the Altered Block Map corresponding to a data block and
including a Modification Bit and a Pending Reset Bit; and the
computer-program product further comprises: means, provided on the
computer-readable medium, for setting the Modification Bit whenever
the data block is modified by a WRITE request; and means, provided
on the computer-readable medium, for setting the Pending Reset Bit
whenever the data block is requested by a client.
25. The computer-program product as in claim 24 further comprising
means, provided on the computer-readable medium, for resetting the
Modification Bit and the Pending Reset Bit upon a request from one
of the clients to the request processor.
26. The computer-program product as in claim 25 further comprising
means, provided on the computer-readable medium, for resetting the
Pending Reset Bit associated with the data block if the data block
is further modified after the respective Pending Reset Bit has been
set and while the Modification Bit remains set.
27. The computer-program product as in claim 25 wherein: the
storage system further includes an Altered Block Counter for
indicating the number of data blocks that have been modified; and
the computer-program product further comprises means, provided on
the computer-readable medium, for incrementing the Altered Block
Counter whenever one of the data blocks is modified.
28. The computer-program product as in claim 27 further comprising
means, provided on the computer-readable medium, for decrementing
the Altered Block Counter whenever the Modification Bit is
reset.
29. The computer-program product as in claim 28 further comprising
means, provided on the computer-readable medium, for decrementing
the Altered Block Counter and means, provided on the
computer-readable medium, for resetting the Pending Reset Bit
whenever the Modification Bit is reset while the Pending Reset Bit
is still on.
30. The computer-program product as in claim 22 further comprising:
means, provided on the computer-readable medium, for the clients to
access the data blocks through a set of application programs; and
means, provided on the computer-readable medium, for the
application programs to communicate with the request processor to
access the desired data blocks and to update the modification
information.
Description
TECHNICAL FIELD
[0001] This invention relates to computer storage systems, and more
particularly to a storage system and method for providing
consistent data modification information to users of the storage
system based on data modification maps.
BACKGROUND OF THE INVENTION
[0002] Data information systems need to store and maintain data and
use storage devices to hold data persistently. New data is
introduced and existing data is modified regularly. Determining
what information has been added to or modified on a storage device
over a specific time interval is necessary for back-up and to
provide redundancy for data stored on the storage device.
Modification information can also be of interest to transactional
systems that are concerned with ensuring that updated data was
updated successfully on a storage device.
[0003] To ensure the availability and integrity of data, data is
often backed up, archived or otherwise replicated. The back-ups,
archives and replicates of information represent a set of critical
functions required by many data information systems.
[0004] Data back-up is typically performed on a per file basis to
allow individual files to be restored. Multiple versions of a file
are usually stored by a back-up system, allowing access to older
versions of a file. However, keeping multiple versions of data
require substantially more storage space than the space occupied by
the data being backed up. The need for more storage space, coupled
with the fact that back-up data is typically not often referenced,
encourages the use of lower cost storage media. If a file becomes
corrupted at a point in time, it is possible that the file may be
restored to a previous version to restore the file's integrity.
[0005] After taking an initial full file system back-up, a common
method for back-up is to determine which files in the file system
have been modified, by examining the file modification stamp to see
if it has changed since the last back-up. If the file was modified,
then the file data is copied. This method is referred to as
incremental back-up. Incremental back-ups reduce the amount of data
that is copied. The file system maintains the modification
information and the file system interface can be used to determine
which files have been modified. Also, since all the file data is
copied, it is easy to collocate file data on the destination
storage medium. This is advantageous for data being written to
sequential media such as tape. A disadvantage to this method is
that if only a portion of a file has been modified, the amount of
data copied may be substantially more than what was modified.
[0006] An alternative method for data back-up is to determine what
portions of a file have been modified and to copy only those
portions that have been modified. One method to accomplish this is
differential back-up. Differential back-up stores a compressed
image of the file. Pieces of the compressed image can be compared
against the file to determine if a portion of the file has been
modified. Differential back-up has proven effective and is
particularly useful for laptop computers or other computing devices
that have limited bandwidth between the device and the destination
storage medium. A disadvantage of this approach is that the host
system has a compressed file image that requires resources to
compress and to store the result. The host needs to examine the
compressed image of modified files to determine what portions of
the file have been modified. While an effective technology in
environments where the rate of data modification is relatively low,
it is less effective in environments where data modification occurs
frequently or on a large amount of data or where host processor
capacity is at a premium.
[0007] In addition to being backed up, data can be replicated to
ensure that it is available from more than one source. Replication
can be performed either dynamically or periodically. Dynamic
replication ensures that replicated data is kept consistent at all
times. Periodic replication ensures that data is guaranteed to be
consistent only at specified times. At other times the device
holding the data to be replicated and the devices that hold copies
of that data may not be fully consistent.
[0008] Mirroring is an example of dynamic data replication. A
storage device has its contents "mirrored" by one or more other
storage devices forming a mirror set. Updates that occur will be
applied simultaneously to each of the mirror set storage devices,
keeping each device's data consistent with the other members of the
mirror set. Mirroring can also be used to make data more widely
available by making it simultaneously accessible from more than one
device. Mirroring ensures that in the event of a device failure
that device's current data remains available. However, mirroring
can increase the latency of updates. The provider of the mirroring
service must also have a mechanism to handle failure events to
ensure that the mirrored devices remain coherent. Maintaining
consistency between members of a mirrored set of devices needs to
occur even during peak workloads, when resources are constrained.
In addition, mirroring cannot be used in place of back-up. A
back-up is still needed for recovering previous versions of a file
or to recover a file if its is inadvertently deleted.
[0009] Replication can also be performed on a periodic or delayed
basis. Periodic replication does not provide instant access to data
in the event of a device failure. Such an approach does not provide
a mirror set that is coherent except for those times when
replication is performed. Data archival is an example of delayed
replication. The contents of an archive are a replica of the data
at some point in time, but changes occurring after the archive was
made are not reflected. Data archival takes a copy of data
off-line. Archived data can be combined with incremental back-ups
to apply modifications to archived data. Data archival is an
expensive process, in that typically all the data from a storage
device is copied with each archival.
[0010] Various methods have been employed to determine what data on
a storage device has been added or modified over some period of
time because copying more data than is necessary for back-up,
archival or replication purposes is undesirable. These methods
typically are external to the storage device itself, often on a
host system that owns the storage device or on a host adapter to
which the storage device is attached. Host systems typically store
data modification information in the form of a modification time
stamp associated with each file within a file system. Storing and
managing modification information on a host system are not an
efficient use of the host's computing resources and result in
poorer overall performance.
[0011] A file system could be implemented to track modifications on
a per block basis. It might accomplish this by storing modification
information about each block in the meta-data that the system keeps
about each file. However, it would be difficult to ensure that file
data and meta-data about the file are consistent with respect to
one another in the event of a system failure. Any host system on
which such a file system would exist would incur the additional
overhead of such a facility.
[0012] Regardless of the method selected, it is desirable to reduce
the overhead associated with determining what information has been
added to or modified on the storage device. The need is exacerbated
as storage devices increase in storage capacity and as more data
needs to be processed.
[0013] Therefore, there remains a need for a method and storage
system that can efficiently provide consistent data modification
information to the clients without the drawbacks described
above.
SUMMARY OF THE INVENTION
[0014] It is an object of the present invention to provide a
storage system and method for presenting consistent data to
multiple clients as the data is regularly updated to assure data
integrity.
[0015] It is another object of the invention to provide consistent
data to the clients based on data modification maps that indicate
the modification status of data blocks and identify data blocks
that were modified.
[0016] It is still another object of the invention to provide
consistent data to the clients using bit maps as the data
modification maps in which the bits respectively represent the
modification status of the data blocks.
[0017] It is a further object of the invention to improve the
recognition of modified data blocks using a Modification Counter
and a Pending Reset Counter.
[0018] To achieve these and other objects, the present invention
provides a storage system that includes a storage device for
storing data and data modification information, a request processor
for handling requests for data from the clients, and operations to
provide consistent data. The request processor is capable of
identifying data that has been modified during a given time
interval based on the modification information and updating the
modification information to ensure that the data accessed by the
clients will be consistent. Preferably, data is stored on the
device as data blocks and the data modification information
includes an Altered Block Map and a Toggle Block Map. The Altered
Block Map indicates the respective modification status of the data
blocks while the Toggle Block Map identifies which data blocks have
been modified over time. The Toggle Block Map is a bit map in which
each bit corresponds to a data block and is complemented whenever
the data block is modified. The Altered Block Map is also a bit map
in which each pair of bits corresponds to a data block and includes
a Modification Bit and a Pending Reset Bit. The Modification Bit is
set whenever the data block is modified by a WRITE request. The
Pending Reset Bit is set whenever the data block is requested by a
client. The Modification Bit and the Pending Reset Bit may be reset
by a request from a client to the request processor. If the data
block is further modified after the respective Pending Reset Bit
has been set, then the Pending Reset Bit is reset while the
Modification Bit remains set.
[0019] The system of the invention further includes an Altered
Block Counter for indicating the number of data blocks that have
been modified and a Modification Counter for indicating whether
data has been modified over a period of time. The Altered Block
Counter is incremented whenever one of the data blocks is modified.
If the Modification Bit is reset while the Pending Reset Bit is
still on, then the Altered Block Counter is decremented and the
Pending Reset Bit is reset. The Modification Counter is incremented
each time one of the data blocks is committed to persistent storage
as a result of a client request.
[0020] The present invention allows clients to access the desired
data blocks and update the modification information through a set
of application programs that submit requests to the request
processor. The invention minimizes the overhead associated with
various forms of data replication by distributing the mechanism to
store and maintain the data modification information to the storage
device. It facilitates rapid data replication and can also be used
to perform incremental back-up and periodic data archival. When
replication is performed, the replica can be used as the data
source to perform back-up and archival operations, eliminating the
need to access the device that was replicated. Additionally, the
storage device of the invention can store coherent data
modification information for use by a transactional facility such
as a database. This information can be examined to ensure that data
written to the storage device was actually committed.
[0021] Additional objects and advantages of the present invention
will be set forth in the description which follows, and in part
will be obvious from the description and with the accompanying
drawing, or may be learned from the practice of this invention.
BRIEF DESCRIPTION OF THE DRAWING
[0022] FIG. 1 is a schematic diagram showing the data storage
system of the invention.
[0023] FIG. 2 illustrates the modification maps for two consecutive
data blocks in a storage device component in accordance with the
invention.
[0024] FIG. 3 is a flow chart showing a preferred embodiment of the
process for updating the Tmap for a single block WRITE request,
according to the invention.
[0025] FIG. 4 is a flow chart showing a preferred embodiment of the
process for updating the Amap for a single block WRITE request,
according to the invention.
[0026] FIG. 5 is a flow chart showing a preferred embodiment of the
process for selecting modified data blocks for processing,
according to the invention.
[0027] FIG. 6 is a flow chart showing a preferred embodiment of the
process for resetting the Amap of a modified data block, according
to the invention.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0028] The invention will be described primarily as a system and
method for providing data consistency in a data storage system
using a request processor and data modification information.
However, persons skilled in the art will recognize that an
apparatus, such as a data processing system, including a CPU,
memory, I/O, program storage, a connecting bus, and other
appropriate components, could be programmed or otherwise designed
to facilitate the practice of the method of the invention. Such a
system would include appropriate program means for executing the
operations of the invention.
[0029] Also, an article of manufacture, such as a pre-recorded disk
or other similar computer program product, for use with a data
processing system, could include a storage medium and program means
recorded thereon for directing the data processing system to
facilitate the practice of the method of the invention. Such
apparatus and articles of manufacture also fall within the spirit
and scope of the invention.
[0030] FIG. 1 illustrates a storage system 100 configured to
perform the operations of the invention. The storage system 100
includes a request processor 103 and one or more storage devices
101. Each device 101 may include one or more hard disk drives,
optical disks, tape drives or other types of data storage medium.
Data is typically stored in the storage devices 104 as data blocks
(or data sectors). All of the blocks on a storage device 101 are
the same size; however, the actual value of the block size does not
depend on the device. For some devices, the block size may be
chosen from a list of the block sizes supported by the devices. In
addition to the data accessed by the clients, the storage devices
104 also contain data modification information needed for the
operations of the invention, as described in detail below in
reference to FIGS. 2-6.
[0031] In accordance with the invention, multiple clients 102 can
store and retrieve data to and from the storage devices 104 through
a request processor 103. The clients 102 accomplish these tasks by
submitting data requests to the request processor 103. There are
five general types of client requests: READ, WRITE, READ MODIFIED
BLOCKS, ACKNOWLEDGE MODIFIED BLOCKS, and OBTAIN TOGGLE MAP
STATE.
[0032] READ requests allow the clients 102 to obtain contents of
the requested data blocks that are stored on the storage devices
104. As READ requests do not alter the state of data or
modification information, they are not discussed further. WRITE
requests update one or more data blocks with data supplied by the
client requests. When a WRITE request "commits" the modified data,
the data update becomes persistent. That is, the modified data
blocks are stored to a persistent storage device 101 and the
modification information for those modified blocks must be updated
and kept in persistent storage. The READ MODIFIED BLOCKS request
allows a client to obtain the location and contents of one or more
modified data blocks. The ACKNOWLEDGE MODIFIED BLOCKS request
provides a mechanism to let the client acknowledge that it
successfully received the modified blocks returned to it as a
result of a previous read modified block request. The OBTAIN TOGGLE
MAP STATE request allows a client to obtain the toggle map value
for one or more specified blocks.
[0033] The modification information on the storage devices 104
allows the clients 102 to recognize when data blocks have been
modified and to quickly determine which blocks have been modified.
In the preferred embodiment of the invention, the modification
information is stored together with client data on the same storage
device 101, however, it can be also maintained in another
persistent storage such as a battery-backed-up RAM or a separate
storage device 101.
[0034] FIGS. 2 depicts a preferred embodiment of the facilities
within a storage device 201 to support the operations of the
invention. To allow the clients 102 to quickly recognize which
blocks of data have been modified, a group of data modification
maps are maintained on the storage device 201. There are two types
of data modification maps: Toggle Block Maps (Tmaps) 202 and
Altered Block Maps (Amaps) 203. While both types of maps maintain
the data modification information, their contents and functions
differ as described below.
[0035] The storage device 201 allows multiple copies of either or
both types of the block modification maps to be defined. The number
of modification maps for each map type on a device is determined
when the administrator initializes the storage device. The clients
102, specifically the administrator, can control the modification
information in several ways. The administrator can select the block
size supported by the storage device and define the number and
types of block modification maps for the storage device. The
administrator can also control the use of the modification maps by
suspending and resuming the monitoring of the data block
modification information by the device.
[0036] In addition to the modification maps, a very small portion
of storage space on the device 201 is used to store control
information and system metadata. This data include configuration
information such as the selected device block size, the number and
types of the modification maps and the name of each map, and any
other data needed by the storage device to correctly function and
to interface with clients. The amount of storage space needed for
the modification information, control information and system
metadata depends on the format of the modification data.
[0037] A client 102 can send requests to the request processor 103
to query, update or control the monitoring of modification
information for the storage device as described in detail below in
reference to FIGS. 3-6. It can query a variety of information such
as the number and types of modification maps, information about
modification maps such as their assigned names, how large they are,
the size of the storage device and the supported device block sizes
and the block size selected, information about which blocks have
been modified. A client 102 can update modification information by
initiating a reset of selected block modification information under
control of the storage device.
[0038] Although the modification maps 202 and 203 might be
implemented in various forms, they will be described as bit maps to
simplify the description of the invention. Other implementations
may provide the block modification information based on the block
numbers of the modified blocks that are stored in a readily
accessible data structure such as a tree or a hashed queue. The
preferred data modification map is an array of bits that store
values representing the modification status for each of the data
blocks on the storage device.
[0039] FIG. 2 illustrates a Tmap 202 and an Amap 203 which reside
respectively in two consecutive data blocks 206 on the storage
device 201. The Tmap 202 preferably includes a bit array with one
Toggle Bit 207 allocated for each block 206 of storage that may be
assigned to hold client data. Whenever a block 206 is modified by a
write, the Toggle Bit 207 associated with that block is
complemented. The operations for processing a possible update to a
block mapped by the Tmap 202 are described below in reference to
FIG. 3.
[0040] The Amap 203 is a bit array in which two bits 204 and 205
are allocated for each block 206 of storage that may be assigned to
hold client data. Collectively, the bits 204 and 205 are known as
the Altered Block Bit pair. Bit 204 of the pair is called the
Modification Bit and is set whenever the associated data block is
modified by a write. The process for updating the Modification Bit
204 when the block is modified is described below in reference to
FIG. 4. The other bit (205) is referred to as a Pending Reset Bit
and is set as the result of a client request. FIG. 6 illustrates
how the Pending Reset Bit 205 gets set for a block when a client
sends the data for that modified block. FIG. 4 illustrates how the
Pending Reset Bit 205 is reset as the result of a WRITE
request.
[0041] FIG. 2 additionally shows an Altered Block Counter 208 and a
Modification Counter 209 on the storage device 201. These counters
help the clients 102 quickly recognize the data on the storage
device that has been modified. A particular embodiment of the
invention may implement either or both of the counters. The Altered
Block Counter 208 is particularly useful when it accompanies the
Amap 203. It is incremented whenever a previously unmodified data
block is modified. A block that was previously modified will have
the Modification Bit 204 already on. If the Modification Bit 204 is
not set, then the Altered Block Counter 208 is incremented,
otherwise (i.e. the bit is already set) the Altered Block Counter
208 remains unchanged. This counter is decremented whenever a data
block that has a Pending Reset Bit set gets reset (i.e. has the
Modification Bit and Pending Reset Bit both cleared as the result
of an ACKNOWLEDGE MODIFIED BLOCKS request).
[0042] The Modification Counter 209 allows the clients 102 to
recognize whether data has been modified over a given interval of
time. Typically, a client 102 would interrogate the value of the
Modification Counter 209 and then does this again at some later
time. If the two values match, no data was modified during the
interval. The request processor 103 increments the Modification
Counter 209 whenever a modified data block is committed to
persistent storage. Preferably, the Modification Counter 209 is a
wrapping monotonically increasing persistent counter.
[0043] While the Altered Block Counter 208 and Modification Counter
209 are optional features, their presence allows for faster and
more efficient recognition and processing of modified data from the
storage device 101. A typical preferred embodiment of the invention
may include just one Modification Counter 209 but multiple Altered
Block Counters 208, one for each Amap 203. An Altered Block Counter
is needed for each Amap 203 because different applications may be
tracking data modification over different intervals of time using
different Amaps.
[0044] A WRITE request can cause a change in the value of the bits
that map a given data block for each modification map. A WRITE to a
given data block will cause the value of the Toggle Bit 207 of the
Tmap 202 that maps that block to toggle (i.e. the value of the bit
is inverted). Further, a WRITE request to a given data block can
cause the Amap 203 bit values to change. A WRITE request causes the
Modification Bit 204 to be set if it was not already set and causes
the Pending Reset Bit 205 to be cleared if that bit was previously
set. Additionally, a write can cause some optional Block Counters
208 and 209 to be incremented.
[0045] To process a WRITE request that modifies a data block, the
request processor 103 first updates or replaces the content of the
data block with client-supplied data. It then commits the modified
data by taking the following actions:
[0046] storing the modified data block to persistent storage
104,
[0047] incrementing the persistent Modification Counter,
[0048] updating any block modification maps and the associated
counters to indicate that the block that was modified, and
[0049] storing the updated maps to persistent storage.
[0050] The data commit ensures that the last three actions above
occur atomically (from the prospective of an outside observer, as a
single indivisible action that occurs completely or not all). The
invention can accomplish this atomicity by the use of a log or
other means of persistent storage that can aggregate the actions
into a single related transaction.
[0051] The request processor 103 processes client requests that
allow data on the storage device 101 to be examined and modified.
It also allows designated clients to examine any data that has been
modified over a given interval of time. Three requests facilitate
clients examining information about whether data on the storage
device was modified: READ MODIFIED BLOCKS request, ACKNOWLEDGE
MODIFIED BLOCKS request and OBTAIN TOGGLE MAP STATE request. The
READ MODIFIED BLOCKS request and the ACKNOWLEDGE MODIFIED BLOCKS
request cause the Amap 203 values for modified data blocks to be
modified under certain conditions. The OBTAIN TOGGLE MAP STATE
request simply allows values in the Tmap 202 to be examined.
[0052] The READ MODIFIED BLOCKS request allows a client to obtain
the location and contents of one or more modified data blocks. In a
preferred embodiment of the invention the request may specify the
maximum number of modified blocks that may be returned as a result
of the request. The ACKNOWLEDGE MODIFIED BLOCKS request provides a
mechanism to let the client acknowledge that it successfully
received the modified blocks returned to it as a result of a
previous read modified block request. In a preferred embodiment of
the invention the READ MODIFIED BLOCKS request and the ACKNOWLEDGE
MODIFIED BLOCKS request may be combined into a single request so
that blocks successfully processed in a previous request may be
acknowledged while at the same time requesting additional modified
blocks be returned if any remain. The OBTAIN TOGGLE MAP STATE
request allows a client to obtain the toggle map value for one or
more specified blocks.
[0053] FIG. 3 illustrates how the Tmap 202 and its toggle bit 207
can be used in conjunction with the OBTAIN TOGGLE MAP STATE request
to determine if a WRITE request successfully completed. FIG. 3
details the processing of how the Toggle Map (Tmap) bits are
processed. When the request processor 103 processes a WRITE request
in step 301, the data block content is modified (or replaced) and
the result is committed in step 302. Additionally, if an Tmap 202
is defined and processing for that Tmap is not suspended (step
303), then the Tmap bit that maps the data block being modified is
checked and toggled in step 305. If the value was set, then it is
reset in step 306. If the value was not set, then it becomes set in
step 307. If processing on the Tmap is suspended, then no further
actions are necessary in step 304. The updated Tmap value is
committed in step 308, completing the update of the data block.
From the perspective of the client that sent the WRITE request 301,
the steps 301-308 appear as an atomic transaction meaning that
steps 302 and 308 are either completed or not completed
together.
[0054] Table 1 illustrates a Tmap 202 for a storage device with
eight blocks available for assignment by clients. The Tmap 202 bits
0 through 7 are used to map the eight data blocks on the storage
device 101. These are the Tmap 202 bits for the data blocks on the
storage device. The value of the Modification Counter 209 is also
shown in the table. Each row in the table represents gives the
contents of the Tmap 202 after the operation specified in the
column Operation Performed has been performed. Note that the
preferred embodiment of the invention illustrated is one that
allows multiple blocks (even blocks that are not logically
adjacent) to be modified in a single WRITE request.
1TABLE 1 Tmap: Blocks 0-7 Modification Bits 0-3 Bits 4-7 Counter
Operation Performed 0000 0000 0 Tmap defined and device initialized
1000 0000 1 Tmap content after a write to block 0 1010 0000 2 Tmap
content after write to block 2 1010 1001 3 Tmap content after
writes to block 4 and 7 0010 1001 4 Tmap content after write to
block 0 1010 0001 5 Tmap content after writes to block 0 and 4
[0055] FIG. 4 illustrates a WRITE request being processed by the
request processor 103 that results in the Amap 203 being updated.
The WRITE request to modify a data block is received in step 401
and the data block content is modified (or replaced) and the result
is committed. If an Amap 203 is defined and processing for that
Amap is not suspended 403 then the Amap's Modification Bit is set
(step 405) and the Pending Reset Bit is reset for the Amap entry
that maps that data block (step 406). If processing on the Amap 203
is suspended then no further actions are necessary (step 404). The
updated Amap 203 value is next committed (step 407), completing the
update of the data block. From the perspective of the client that
sent the WRITE request (step 401), the steps 401-407 appear as an
atomic transaction meaning that steps 402 and 407 are both either
completed or not completed.
[0056] FIG. 5 illustrates the processing related to a READ MODIFIED
BLOCKS request. The illustration represents a preferred embodiment
of the invention where the request processor 103 receives a
combined request to READ MODIFIED BLOCKS and to ACKNOWLEDGE
MODIFIED BLOCKS (from a previous request) in step 501. The
processing described focuses on how the READ MODIFIED BLOCKS
portion of the request is processed. The request specifies the Amap
203 the request is against and a starting offset in the Amap 203 to
begin the search as well as a maximum block count to be returned
for modified blocks (step 502). The request processor 103 finds the
next modified block (the next block that has it's Modification Bit
set). It sets the Pending Reset Bit of the Amap for that block in
step 503 and prepares to return the contents of the modified data
block to the request processor in step 504. The updated Amap value
is committed (step 505) and the modified data block information
along with their location on the device and a count of blocks being
returned is sent back to the requesting client (step 506).
[0057] FIG. 6 illustrates how an ACKNOWLEDGE MODIFIED BLOCKS
request provides a mechanism to let the client acknowledge that it
received and processed data blocks returned to it from a previous
READ MODIFIED BLOCKS request. The illustration represents a
preferred embodiment of the invention where the request processor
103 receives a combined request to READ MODIFIED BLOCKS and to
ACKNOWLEDGE MODIFIED BLOCKS (from a previous request) in step 601.
The processing described focuses on how the ACKNOWLEDGE MODIFIED
BLOCKS portion of the request is processed. In this illustration,
the request specifies one (or more) data block that was returned to
the client from a previous READ MODIFIED BLOCKS request and was
successfully processed by the client (step 602). The client is
informing the request processor 103 that from the perspective of
the client, the client believes it has examined the most up to date
content for this data block. In reality, this may not be the case
if the data block was further modified. The request processor 103
checks the Pending Reset Bit 205 of the Amap that maps this data
block in step 603. If the Pending Reset Bit is set the block has
not been further modified so the acknowledge modified block request
for that block is successful. A successful acknowledge modified
block request for a block results in the Amap's Modification Bit
204 to be reset in step 605, along with the Amap 203's Pending
Reset Bit 205 being reset in step 606. The updated Amap value is
committed in step 607 and the result of the request is returned to
the client in step 608. If the Amap's Pending Reset Bit for the
data block is not set (step 603), then no further processing is
needed (step 604). This results in the client receiving a result
that their acknowledge modified block request failed (informing
them that the data block has been further modified).
[0058] It is important to note that if a WRITE to given data block
intervenes between READ MODIFIED BLOCKS request that set the
Pending Reset Bit for that block and the ACKNOWLEDGE MODIFIED
BLOCKS request for the block fails, so the Modification Bit for
that block remains set. This is critical to maintaining the correct
state of modified data blocks as the block in question has been
further modified.
[0059] The Altered Block Counter if present is also affected by
reset processing. Again, the Altered Block Counter is incremented
whenever a data block that does not have its Modification Bit set
has that bit set (i.e. a WRITE request modifies the block). The
Altered Block Counter is decremented when ever a data block that
has a set Pending Reset Bit gets reset (i.e. due to an ACKNOWLEDGE
MODIFIED BLOCKS request).
[0060] Table 2 illustrates a successful processing of the Amap. The
example includes an Amap 203 for a storage device with eight blocks
available for assignment by clients. The Amap bits 0 through 15 are
used to map the storage device's eight data blocks. Two bits, a
Modification Bit 204 and Pending Reset Bit 205 are used for each
block mapped by the Amap 203. The values for the Altered Block
Counter 208 and the Modification Counter 209 are also shown in the
table. Each row in the table represents gives the contents of the
Amap 203 after the operation specified in the column Operation
Performed has been performed. Note that the preferred embodiment of
the invention illustrated allows multiple blocks (even blocks that
are not logically adjacent) to be modified in a single WRITE
request. Note also that in a preferred embodiment of the invention
the steps illustrated in the READ MODIFIED BLOCKS request and
ACKNOWLEDGE MODIFIED BLOCKS request could be combined into a single
request. However in Table 2, the two requests are separated for the
sake of clarity.
2TABLE 2 Modifi- Altered Amap: Blocks 0-7 cation Block Bits 0-7,
8-15 Counter Operation Performed Counter 0000 0000 0000 0000 0 Amap
defined and device 0 initialized 1000 0000 0000 0000 1 Amap content
after a 1 write to block 0 1000 1000 0000 0000 2 Amap content after
2 write to block 2 1000 1000 1000 0010 3 Amap content after 4
writes to blocks 4 and 7 1100 1100 1000 0010 3 Amap content after 4
READ MODIFIED BLOCKS request - two blocks returned 0 and 2 0000
0000 1000 0010 3 Amap content after 2 a successful ACKNOWLEDGE
MODIFIED BLOCKS request for blocks 0 and 2. 0000 0000 1100 0011 3
Amap content after 2 READ MODIFIED BLOCKS request - two blocks
returned 4 and 7 0000 0010 1100 0011 4 Amap content after 3 write
to block 3 0000 0010 0000 0000 4 Amap content after 1 ACKNOWLEDGE
MODIFIED BLOCKS request for block 4 and 7.
[0061] Table 3 illustrates the processing of an Amap 203 with an
intervening WRITE occurring between a READ MODIFIED BLOCKS request
and a subsequent ACKNOWLEDGE MODIFIED BLOCKS request. Again, the
Amap 203 has eight entries mapping the hypothetical eight blocks on
the storage device, where two bits are assigned for each entry: one
Modification Bit 204 and one Pending Reset Bit 205. Each row in the
table represents gives the contents of the Amap 203 after the
operation specified in the column Operation Performed has been
performed. The same comments about the preferred embodiments
mentioned for Table 2 apply. The difference in this example is that
intervening WRITE occurs just prior to the final acknowledge modify
blocks request occurring. This negates the reset for block four as
that block has been further altered since the block was marked for
reset. The intervening WRITE cleared the Pending Reset Bit so the
block is easily recognized as being altered and so it remains
marked as modified. The Altered Block Counter is also not
decremented since the block is still modified.
3TABLE 3 Modifi- Altered Amap: Blocks 0-7 cation Block Bits 0-7,
8-15 Counter Operation Performed Counter 0000 0000 0000 0000 0 Amap
defined and device 0 initialized 1000 0000 0000 0000 1 Amap content
after a 1 2 Amap content after 2 a write to block 2 1000 1000 1000
0010 3 Amap content after 4 writes to blocks 4 and 7 1100 1100 1000
0010 3 Amap content after 4 READ MODIFIED BLOCKS request - two
blocks returned 0 and 2 0000 0000 1000 0010 3 Amap content after a
2 successful ACKNOWLEDGE MODIFIED BLOCKS request for blocks 0 and
2. 0000 0000 1100 0011 3 Amap content after 2 READ MODIFIED BLOCKS
request - two blocks returned 4 and 7 0000 0010 1000 0011 4 Amap
after blocks 3 3 and 4 are modified 0000 0010 1000 0000 4 Amap
content after 2 ACKNOWLEDGE MODIFIED BLOCKS request for block 4 and
7. Note that block 3 was further modified so it is sill marked
modified.
[0062] While the present invention has been particularly shown and
described with reference to the preferred embodiments, it will be
understood by those skilled in the art that various changes in form
and detail may be made without departing from the spirit and scope
of the invention. Accordingly, the disclosed invention is to be
considered merely as illustrative and limited in scope only as
specified in the appended claims.
* * * * *